Early depression detection using ensemble machine learning framework

Social media platforms typically serve as generators of huge data sources as users express their sentiments directly or indirectly on these platforms. With increased sentiment data on social media platforms, there is an immediate requirement to design intelligent systems with early risk detection (ERD)capabilities. Early detection in case of mental health disorders, especially in case of depression detection could provide for better information, identification, and utilization of treatments, along with future risk reduction planning. Machine learning based early risk detection systems typically utilize the social media sentiments to correctly classify the potential depression cases to initiate an early diagnosis, detection, and recovery process. The authors, henceforth, present a unique ensemble-based machine learning classifier with a mix of logistic regression, decision tree, random forest, support vector machine, multi- layer perceptron along with adaptive and gradient boosting resulting in improved performance on evaluation metrices indicators than past research.